OF IRON, COPPER, AND GERMAN SILVER. 557 



It only now remains to compare briefly the results given above with 

 previous work upon the same bars. The particular values of the rates of 

 cooling, and also of the tangents to the curve of stationary temperature excess, 

 which Forbes used, have been already alluded to. Herein, it would seem, lies 

 the explanation of his result that the thermometric conductivity of iron 

 decreases as temperature increases. Employing, as he did, values of dvjdx, 10 

 per cent, larger than those found in this work, and also values of the rate of 

 cooling which are now known to be much too small at any temperature excess 

 beyond 150° C, these two factors are quite sufficient to account for the large 

 decrease in conductivity which he stated for iron. 



These remarks are fully confirmed by the results of Prof. Tait, which are 

 intermediate between those of Forbes and those given by me in the tables 

 above. In Diagram 1 it is shown that the rates of change of temperature excess 

 at different points along the long bar were, for the curve representing the experi- 

 ments in question, intermediate between those of Forbes and my own. Again, 

 in regard to the rates of cooling, the same may be said, as has already been 

 referred to in connection with Diagram 6. This explains why Prof. Tait's results 

 for iron (viz., that its thermal conductivity is as nearly as possible constant), 

 should lie between those of Forbes (already stated) and those given in this 

 paper, where the conductivity of iron is shown to increase ; for, because of 

 reasons already given, the results derived from the experiments on the iron bat- 

 cooled midway, are certainly more trustworthy than those of the ordinary form 

 of the experiment. 



With respect to the copper and German silver bars, the results for these 

 differ mainly from those of Prof. Tait in regard to the rate at which the 

 conductivity rises. In both samples of copper the rise is less, but in German 

 silver more than in the previous work on these bars. But as to the absolute 

 value of the conductivity there is no great difference between the two series of 

 estimates. 



In conclusion, I have to tender thanks to Prof. Tait for the interest he has 

 shown throughout in this inquiry, and for much valuable assistance and 

 guidance both in the experimental part of the work and in the more laborious 

 work of reduction. I have also to thank my friend Mr Robert Dickinson for 

 the care he has bestowed upon the accurate execution of the diagrams 

 published along with this paper. 



